3.3 Continuous and Discontinuous Measurement Systems

Continuous and Discontinuous Measurement Systems

Accurate, reliable data collection is essential for assessing behavior and evaluating treatment efficacy. Qualified Behavior Analysts must select measurement systems that match the specific dimensional quantities of the target behavior while accounting for observational resources in clinical settings. Measurement systems are broadly classified into two categories: continuous measurement (recording every instance of behavior) and discontinuous measurement (sampling behavior across specified time intervals).

Fundamental Dimensions of Behavior

All behavioral measures derive from three basic dimensional quantities defined by Johnston and Pennypacker:

• Repeatability (Countability): Instances of a response class can occur repeatedly through time (Count, Rate/Frequency).
• Temporal Extent: A response occupies an explicit duration of time (Duration).
• Temporal Locus: A response occurs at a specific point in time relative to other events (Latency, Interresponse Time).

---

Continuous Measurement Systems

Continuous measurement systems capture all occurrences of a target behavior during an observation window. These procedures provide the most accurate portrayal of behavior but require undivided observer attention.

1. Count and Rate/Frequency

• Count: The simple tally of the number of occurrences of a behavior. Useful only when observation session lengths are strictly constant.
• Rate (Frequency): The ratio of count per unit of time.

$\text{Rate} = \frac{\text{Total Count}}{\text{Total Observation Time}}$

• Clinical Application: Expressed as responses per minute, hour, or day. Rate is essential when observation session lengths vary across days.
• Contraindications: Do not use rate for behaviors that occur within discrete trials (where opportunities are constrained by the practitioner) or for behaviors with prolonged, variable durations (e.g., sustained crying or academic engagement).

2. Duration

Duration measures the total amount of time from the onset of a behavior to its offset.

• Total Duration per Session: Cumulative sum of time engaged in the target behavior across an entire session.
• Duration per Occurrence: The measure of duration for each individual instance of the target behavior.
• Clinical Application: Used for behaviors characterized by temporal extent, such as task engagement, sleeping, tantrums, or social interaction.

3. Latency and Interresponse Time (IRT)

Both latency and IRT measure temporal locus, but they differ regarding their reference anchor points:

Antecedent Stimulus Onset ----( LATENCY )----> Response 1 Onset
Response 1 Offset ------------(  IRT  )------------> Response 2 Onset

• Latency: The elapsed time between the onset of an antecedent stimulus (e.g., an instruction or demand) and the initiation of the target response. Used when increasing instructional compliance or reducing delayed processing.
• Interresponse Time (IRT): The elapsed time between the offset of one response and the onset of the subsequent response in the same response class.

Mathematical Relationship Between IRT and Rate

IRT and Rate share an inverse mathematical relationship: as average IRT increases, the overall rate of behavior decreases. Conversely, as average IRT decreases, rate increases. QBAs target IRT when implementing Differential Reinforcement of Low Rates (DRL, to increase IRT) or Differential Reinforcement of High Rates (DRH, to decrease IRT).

$\text{Rate} = \frac{1}{\text{Mean IRT}}$

---

Discontinuous Measurement Systems (Interval Sampling)

When continuous measurement is unfeasible due to high-rate behaviors, indistinct response boundaries, or multi-tasking clinicians, behavior analysts utilize discontinuous (time sampling) measurement. The observation period is divided into brief, equal time intervals (e.g., 10 seconds).

---

Decision Matrix for Measurement System Selection

To select the optimal measurement system, QBAs evaluate behavior topographies and clinical goals using the following decision matrix:

Does the behavior have distinct start/stop boundaries and brief duration?
 ├── YES ==> Are observation session lengths variable?
 │           ├── YES ==> Rate / Frequency
 │           └── NO  ==> Count / Frequency
 └── NO  ==> Is continuous tracking of time possible?
             ├── YES ==> Duration, Latency, or IRT
             └── NO  ==> Select Interval Sampling Method:
                         ├── Goal is BEHAVIOR REDUCTION ==> Partial-Interval Recording
                         ├── Goal is SKILL ACQUISITION  ==> Whole-Interval Recording
                         └── Observer has MULTIPLE TASKS ==> Momentary Time Sampling / PLACHECK